Reasoning with uncertainty
Cognitive tension under fundamental uncertainty
The twofold nature of uncertainty (ontological and epistemic) calls attention to the twofold cognitive structure that may be needed in order to cope with its demands. Ontological uncertainty arises from the nature of the world but cannot be expressed except by means of adequate categories. Epistemic uncertainty arises from the cognitive boundaries of human understanding, but is often intermingled with difficulty arising from structural change in the world of objects and situations. Clearly, coping with uncertainty presupposes at the same time representational and inferential abilities, and reasoning with fundamental uncertainty cannot take place unless one is prepared to move back and forth between alternative circumscriptions and alternative ways to single out similarity and likelihood.
The above argument suggests that there is an inherent tension between what may be required for successful handling of ontological uncertainty (that is, how best to circumscribe a changing configuration of objects or situations) and what may be needed for effective handling of epistemic uncertainty (that is, how best to generate the categories needed for that circumscription). A possible way out of this tension is to explicitly acknowledge that fundamental uncertainty is unavoidably associated with the intermingling of the ontological and the epistemic components, so that reasoning with fundamental uncertainty would require both the ability to direct attention to changing circumscriptions of objects and situations, and the ability to conceive of the categories that may be needed in order to express precisely those circumscriptions.20 In particular, reasoning with uncertainty appears to be closely associated with the ability to consider manifold orders of likelihood, so that the likelihood of any given situation would ultimately be assessed after assigning certain 'weights' to the different orders of likelihood relevant for the case in view. For example, the distinction between short-term and long-term likelihood orders suggest a weighting of short-term and long-terms features of similarity (see s. 5.4).